Design and methods in a multi-center case-control interview study

Patients with early arthritis consume less alcohol than controls, regardless of the type of arthritis

OBJECTIVES

There are conflicting reports concerning the association between alcohol consumption and RA. We performed a case-control study to investigate the association of alcohol consumption with RA as well as with other forms of arthritis. To assess whether alcohol consumption affects long-term disease outcome, we also investigated its association with radiographic progression and sustained drug-free remission in RA.

METHODS

Patients with arthritis and various diagnoses including RA, OA, ReA, SpA and PsA were compared with 5868 controls from the general population. The association of disease with alcohol consumption was analysed by logistic regression analysis.

RESULTS

Alcohol consumption was inversely associated with not only RA [odds ratio (OR) 0.28, 95% CI 0.23, 0.35] but also OA (OR 0.31, 95% CI 0.16, 0.62) and other forms of arthritis (OR 0.34, 95% CI 0.24, 0.48). A higher degree of systemic inflammation, reflected by the ESR and CRP level, was associated with a smaller proportion of patients consuming alcohol. There was no dose-response relationship between the amount of alcohol consumed and the presence of arthritis. The extent of joint destruction and the rate of sustained drug-free remission were not affected by alcohol consumption.

CONCLUSION

Arthritis patients report less alcohol consumption than controls, regardless of the type of arthritis. This suggests that alcohol may either protect against different kinds of arthritis or that the inverse association between alcohol and arthritis may be secondary to disease development, with arthritis patients being less inclined to consume alcohol due to their decreased general well-being.

Fruit, vegetables, fibre and micronutrients and risk of US renal cell carcinoma

The association between renal cell cancer (RCC) and intake of fruit, vegetables and nutrients was examined in a population-based case-control study of 323 cases and 1827 controls; dietary intake was obtained using a mailed questionnaire. Cancer risks were estimated by OR and 95 % CI, adjusting for age, sex, smoking, obesity, hypertension, proxy status, alcohol consumption and dietary fat intake and energy. Intake of vegetables was associated with a decreased risk of RCC (OR 0·5; 95 % CI 0·3, 0·7; P trend = 0·002), (top compared to the bottom quartile of intake). When intake of individual nutrients was investigated, vegetable fibre intake was associated with decreased risks (OR 0·4; 95 % CI 0·2, 0·6; P < 0·001), but this was not the case with fruit fibre (OR 0·7; 95 % CI 0·4, 1·1) or grain fibre (OR 1·0; 95 % CI 0·6, 1·5). β-Cryptoxanthin and lycopene were also associated with decreased risks, but when both were included in a mutually adjusted backwards stepwise regression model, only β-cryptoxanthin remained significant (OR 0·5; 95 % CI 0·3, 0·8). When other micronutrients and types of fibre were investigated together, only vegetable fibre and β-cryptoxanthin had significant trends (P < 0·01) (OR 0·6; 95 % CI 0·3, 0·9) (OR 0·5; 95 % CI 0·3, 0·9), respectively. These findings were stronger in those aged over 65 years (P interaction = 0·001). Among non-smokers, low intake of cruciferous vegetables and fruit fibre was also associated with increased risk of RCC (P interaction = 0·03); similar inverse associations were found for β-cryptoxanthin, lycopene and vitamin C. When nutrients were mutually adjusted by backwards regression in these subgroups, only β-cryptoxanthin remained associated with lower RCC risk. These findings deserve further investigation in ongoing prospective studies when sample size becomes sufficient.

The "Got D'ViBE?" study: an inter-institutional project assessing vitamin D and mammographic breast density

A multi-institutional collaboration was forged to implement a study of the relationship between Vitamin D and breast density among medically underserved women. This effort resulted in techniques to measure vitamin D levels, breast density, and sunlight exposure. Outcomes from this collaboration may provide insight to researchers conducting similar investigations.

Dietary fat and risk of renal cell carcinoma in the USA: a case-control study

An increased risk of renal cell carcinoma (RCC) has been linked with obesity. However, there is limited information about the contribution of dietary fat and fat-related food groups to RCC risk. A population-based case-control study of 406 cases and 2434 controls aged 40-85 years was conducted in Iowa (1986-89). For 323 cases and 1820 controls from the present study, information on dietary intake from foods high in fat nutrients and other lifestyle factors was obtained using a mailed questionnaire. Cancer risks were estimated by OR and 95 % CI, adjusting for age, sex, smoking, obesity, hypertension, physical activity, alcohol and vegetable intake and tea and coffee consumption. In all nutrient analyses, energy density estimates were used. Dietary nutrient intake of animal fat, saturated fat, oleic acid and cholesterol was associated with an elevated risk of RCC (OR = 1.9, 95 % CI 1.3, 2.9, P trend < 0.001; OR = 2.6, 95 % CI 1.6, 4.0, P trend < 0.001; OR = 1.9, 95 % CI 1.2, 2.9, P trend = 0.01; OR = 1.9, 95 % CI 1.3, 2.8, P trend = 0.006, respectively, for the top quartile compared with the bottom quartile of intake). Increased risks were also associated with high-fat spreads, red and cured meats and dairy products (OR = 2.0, 95 % CI 1.4, 3.0, P trend = 0.001; OR = 1.7, 95 % CI 1.0, 2.2, P trend = 0.01; OR = 1.8, 95 % CI 1.2, 2.7, P trend = 0.02; OR = 1.6, 95 % CI 1.1, 2.3, P trend = 0.02, respectively). In both the food groups and nutrients, there was a significant dose-response with increased intake. Our data also indicated that the association of RCC with high-fat spreads may be stronger among individuals with hypertension. These findings deserve further investigation in prospective studies.

Tobacco smoking by occupation in Australia and the United States: a review of national surveys conducted between 1970 and 2005

Tobacco use represents a key issue in workplace health, and much can be learned from countries where occupational smoking data has been regularly collected on a national basis. This article reviews national smoking surveys previously conducted in Australia and the United States between 1970 and 2005, and examines their relevance to the field of occupational tobacco control. When the first preliminary research was undertaken over 35 years ago, tobacco smoking was a regular feature in the general population of both countries, albeit with higher rates often documented among blue collar workers. Recent national investigations however, suggest that certain historical differences in smoking prevalence rates by occupation are persisting as we enter the 21st century. Detailed examination also indicates that employee sub-groups, such as cleaners and construction workers, are now bearing much of the occupational smoking burden in Australia and the United States. As such, there is an urgent need for more aggressive and finely targeted tobacco control activities in the workplace, as well as increased cooperation between tobacco control organizations, labor unions and other stakeholders, so that they may more effectively combat this ongoing threat to workers' health.

Interviewer variability - quality aspects in a case-control study

Quality assurance and quality control are important for the reliability of case-control studies. Here we describe the procedures used in a previously published study, with emphasis on interviewer variability. To evaluate risk factors for acute pancreatitis, information including previous diagnoses and medication was collected from medical records and by telephone interviews from 462 cases and 1781 controls. Quality assurance procedures included education and training of interviewers and data validity checks. Quality control included a classification test, annual test interviews, expert case validation, and database validation. We found pronounced variations between interviewers. The maximal number of interviews per day varied from 3 to 9. The adjusted average (95% CI) number of diagnoses captured per interview of cases was 4.1 (3.8-4.3) and of controls 3.5 (3.4-3.7) (excluding one deviating interviewer). For drugs, the average (95% CI) number per interview was 3.9 (3.7-4.1) for cases and 3.3 (3.2-3.4) for controls (excluding one deviating interviewer). One of the fourteen interviewers deviated significantly from the others, and more so for controls than for cases. This interviewer's data ;were excluded. Nonetheless, data concerning controls more frequently needed correction and supplementation than for cases. Erroneous coding of diagnoses and medication was also more frequent among controls. Thus, a system for quality control of coding practices is crucial. Variability in interviewers' ability to ascertain information is a possible source of bias in interview-based case-control studies when "blinding" cannot be achieved.

Body mass index, physical activity, and risk of renal cell carcinoma

OBJECTIVE

To investigate the association between obesity and risk of renal cell carcinoma and to examine whether the association is modified by physical activity.

SUBJECTS

A population-based case-control study of 406 patients with renal cell carcinoma and 2434 controls conducted in Iowa.

METHODS

Information was collected on weight at the ages 20-29, 40-49, and 60-69 years, height, nonoccupational physical activity, diet, and other lifestyle factors. Renal cell carcinoma risk was estimated by odds ratios (ORs) and 95% confidence intervals (CIs), adjusting for age, total energy intake, and other confounding factors.

RESULTS

Height and total energy intake were not associated with risk in either sex. In men, neither physical activity nor level of obesity in any period of life was significantly associated with risk. In women, lower physical activity was associated with higher risk (OR=2.5; 95% CI=1.2-5.2 comparing exercise <1 time/month to >1 time/day). Compared with women in the lowest quartile for BMI, the risks of renal cell carcinoma for women in the highest 10% of BMI in their 20s, 40s, and 60s were 1.4 (CI=0.6-3.1), 1.9 (CI=0.9-4.2), and 2.3 (CI=0.9-6.0), respectively. When analyses were limited to self-respondent data, the corresponding ORs were 2.9 (CI=1.2-7.4), 3.2 (CI=1.3-7.5), and 2.1 (CI=0.7-6.4), respectively. There was little evidence that physical activity modifies the association of BMI with renal cell carcinoma.

CONCLUSION

Nonoccupational physical activity was inversely associated and obesity was positively associated with risk of renal cell carcinoma among women. The risk appeared to be greater for women in the highest 10% of BMI in their 40s. Our finding of little evidence of an interaction between physical activity and BMI requires confirmation.

Beauty product-related exposures and childhood brain tumors in seven countries: results from the SEARCH International Brain Tumor Study

Data from 1218 cases of childhood brain tumors (CBT) diagnosed between 1976 and 1994 and 2223 matched controls from the general population were included in an analysis of maternal beauty product exposure and beauty-related employment in 9 centers in 7 countries. A 50% increased odds ratio (OR) [95% confidence interval (CI) = 1.0-2.1] for CBT was observed among children of mothers who were exposed via personal use of and/or possible ambient contact with beauty products during the 5 years preceding the index child's birth compared with children of mothers never exposed to beauty products during this time period. Overall maternal personal use of hair-coloring agents in the month before or during the pregnancy of the index child's birth was not associated with CBT (OR = 1.0, CI = 0.83-1.3) or with astroglial (OR = 1.1, CI = 0.85-1.4), PNET (OR = 1.0, CI = 0.71-1.5) and other glial subtypes (OR = 1.0, CI = 0.62-1.0). Similarly, no statistically increased ORs or discernable pattern of risk estimates were observed for period of use or for number of applications per year for maternal personal use of hair-coloring agents overall or by histologic type. Among children born on or after 1980, increased ORs for CBT were associated with maternal non-work-related exposure to any beauty products (OR = 2.6, CI = 1.2-5.9), hair-dyes (OR = 11, CI = 1.2-90), and hair sprays (OR = 3.4, CI = 1.0-11). No overall increased OR for CBT was observed among children of mothers employed in beauty-related jobs during the 5 years preceding the index child's birth compared with those who reported no beauty-related employment. In general, other specific beauty product-related exposures were not associated with increased ORs for CBT. Data from our study provide little evidence of an increased risk for CBT with mothers' exposures to beauty products.

Farm-related exposures and childhood brain tumours in seven countries: results from the SEARCH International Brain Tumour Study

A total of 1218 cases of childhood brain tumours (CBT) and 2223 control subjects from the general population were included in a population-based case-control study conducted in nine centres in seven countries. Mothers were asked about farm- or agriculture-related exposures. Significantly elevated odds ratios (OR) for CBT were associated with children's personal and maternal prenatal exposure while living on a farm with pigs (child OR = 1.7, mother OR = 2.3), horses (child OR = 1.6, mother OR = 1.8), dogs (child OR = 1.5, mother OR = 1.5) and cats (child OR = 1.5, mother OR = 1.7). Children who were exposed to pigs, horses and cats combined, while living on a farm, had a threefold elevated OR for CBT. Increased ORs for primitive neuroectodermal tumours (PNET) were associated with children's farm exposure to dogs (OR = 1.9) and cats (OR = 2.2), and maternal farm exposure to pigs (OR = 4.2). The OR for CBT was elevated (OR = 2.3) for children of mothers who had preconception/prenatal farm- or agriculture-related employment involving potential contact with animals, relative to no farm- or agriculture-related employment. In particular, increased ORs for CBT were observed for children of mothers who were employed as general farmers (OR = 4.1) or general farm workers (OR = 3.8). During the 5 years preceding the index child's birth, maternal exposures were related to CBT, relative to no maternal exposure to agricultural chemicals or animal products: fertilisers (OR = 1.8), pesticides (OR = 2.0), animal manure (OR = 2.0) and unprocessed wool (OR = 3.0). Our findings suggest that various farm-related exposures are positively associated with CBT and warrant further investigation into the public health importance of these associations.

Follow-up methods for retrospective cohort studies in New Zealand

OBJECTIVES

To define a general methodology for maximising the success of follow-up processes for retrospective cohort studies in New Zealand, and to illustrate an approach to developing country-specific follow-up methodologies.

METHODS

We recently conducted a cohort study of mortality and cancer incidence in New Zealand professional fire fighters. A number of methods were used to trace vital status, including matching with records of the New Zealand Health Information Service (NZHIS), pension records of Work and Income New Zealand (WINZ), and electronic electoral rolls. Non-electronic methods included use of paper electoral rolls and the records of the Registrar of Births Deaths and Marriages.

RESULTS

95% of the theoretical person-years of follow-up of the cohort were traced using these methods. In terms of numbers of cohort members traced to end of follow-up, the most useful tracing methods were fire fighter employment records, the NZHIS, WINZ, and the electronic electoral rolls.

CONCLUSIONS

The follow-up process used for the cohort study was highly successful. On the basis of this experience, we propose a generic, but flexible, model for follow-up of retrospective cohort studies in New Zealand. Similar models could be constructed for other countries.

IMPLICATIONS

Successful follow-up of cohort studies is possible in New Zealand using established methods. This should encourage the use of cohort studies for the investigation of epidemiological issues. Similar models for follow-up processes could be constructed for other countries.

Occupation and bladder cancer: a population-based, case-control study in Iowa

While considerable efforts have been made to investigate the role of occupation and industry in the risk of bladder cancer, many reported associations have not been consistent, and strong evidence of increased risk is apparent for few occupational groups. To further examine the issue, a large, population-based, case-control study was conducted in the state of Iowa among both men and women. A total of 1452 incident bladder cancer cases and 2434 controls were included in the study. Occupational history was collected from respondents for each job held for 5 years or longer since age 16. Among men, excess risk was observed for industries including plumbing, heating, and air conditioning (odds ratio [OR], = 2.2; 95% confidence interval [CI], 1.0 to 5.0); rubber and plastic products (OR = 3.1; 95% CI, 1.2 to 8.5), motor vehicle parts and supplies (OR = 4.5; 95% CI, 1.2 to 16.5), and occupations including supervisors for transportation and material moving (OR = 6.5; 95% CI, 1.4 to 29.9), material-moving-equipment operators (OR = 1.9; 95% CI, 1.0 to 3.6), automobile mechanics (OR = 1.6; 95% CI, 1.0 to 2.6), painters (OR = 2.7; 95% CI, 1.0 to 7.7), and metal- and plastic-working machine operators (OR = 2.0; 95% CI, 1.1 to 3.4). Among women, significant excess risk was observed for secondary school teachers and record clerks. Housekeepers and butlers and workers in laundering and dry cleaning were also at increased risk. In conclusion, these results suggest that occupational exposures may play a significant role in the risk of bladder cancer.

The NICHD Women's Contraceptive and Reproductive Experiences Study: methods and operational results

PURPOSE

This paper presents methods and operational results of a population-based case-control study examining the effects of oral contraceptive use on breast cancer risk among white and black women aged 35-64 years in five U.S. locations.

METHODS

Cases were women newly diagnosed with breast cancer during July 1994 through April 1998. Controls were identified through random digit dialing (RDD) using unclustered sampling with automated elimination of nonworking numbers. Sampling was density-based, with oversampling of black women. In-person interviews were conducted from August 1994 through December 1998. Blood samples were obtained from subsets of cases and controls, and tissue samples were obtained from subsets of cases. A computerized system tracked subjects through study activities. Special attention was devoted to minimizing exposure misclassification, because any exposure-disease associations were expected to be small.

RESULTS

An estimated 82% of households were screened successfully through RDD. Interviews were completed for 4575 cases (2953 whites; 1622 blacks) and 4682 controls (3021 whites; 1661 blacks). Interview response rates for cases and controls were 76.5% and 78.6%, respectively, with lower rates for black women and older women.

CONCLUSIONS

The methodologic details of this large collaboration may assist researchers conducting similar investigations.

Driver's license and voter registration lists as population-based sampling frames for rural African Americans

PURPOSE

To compare coverage of a state driver's license list and county voter registration lists as frames for sampling rural African Americans.

METHODS

Name, birth date, sex, and county were extracted from records for all 18--59 year-old African Americans residing in eight rural North Carolina counties and listed in the North Carolina Driver license file [obtained as a SAS dataset from the University of North Carolina (UNC) Highway Safety Research Center] and in machine-readable text files of registered voters (obtained from county boards of elections). Files were edited, merged, and matched by last name and date of birth to produce three files for each five-year age group, sex, and county: 1) persons listed only in the driver's file; 2) persons listed only in the voter's file; and 3) persons listed in both.

RESULTS

The median percentages of unique persons found only on the driver's list, only on the voter's list, and on both lists were 54%, 17%, and 30% men, and 35%, 25%, and 41% women, respectively.

CONCLUSIONS

The driver's list provided better coverage than did voter registration lists. Federal legislation that prohibits states from releasing driver's license lists for use in surveys removes a valuable resource for population-based research.

How to find Chinese research participants: use of a phonologically based surname search method

The frequent use of non-representative, time-consuming and low hit rate sampling methods in the study of minority cultures is of concern. Given the increasing cultural diversity of the Canadian population and the growing Chinese population, a sampling method using phonological rules to distinguish Chinese surnames from non-Chinese ones was developed. Chinese surnames, transcribed according to their pronunciations, follow specific phonological rules. A flowchart was developed using the inclusion and exclusion criteria derived from these rules. Methods to validate Chinese surnames (identified from telephone directory) were external reference source, expert panel, and telephone interview. A total of 266 possible Chinese surnames were selected using the flowchart. External reference source validated 153 surnames, expert panel 53, and telephone interview 23. An overall hit rate of 86.1% (229/266) was achieved, reflecting the validity of the phonological criteria in identifying individuals with a Chinese surname from the general population.

How to find Chinese research participants: use of a phonologically based surname search method

The frequent use of non-representative, time-consuming and low hit rate sampling methods in the study of minority cultures is of concern. Given the increasing cultural diversity of the Canadian population and the growing Chinese population, a sampling method using phonological rules to distinguish Chinese surnames from non-Chinese ones was developed. Chinese surnames, transcribed according to their pronunciations, follow specific phonological rules. A flowchart was developed using the inclusion and exclusion criteria derived from these rules. Methods to validate Chinese surnames (identified from telephone directory) were external reference source, expert panel, and telephone interview. A total of 266 possible Chinese surnames were selected using the flowchart. External reference source validated 153 surnames, expert panel 53, and telephone interview 23. An overall hit rate of 86.1% (229/266) was achieved, reflecting the validity of the phonological criteria in identifying individuals with a Chinese surname from the general population.

Cigarette smoking and risk of bladder, pancreas, kidney, and colorectal cancers in Iowa

PURPOSE

Although there are numerous reports on the effects of cigarette smoking and cancer, they have infrequently compared risks at more than one cancer site after multivariate adjustment. We analyzed data from a population-based case-control study that included five anatomic sites to evaluate the association between cigarette smoking and each cancer site and to rank the associations by site.

METHODS

Study respondents included 1452 bladder, 406 kidney, 376 pancreatic, 685 colon, and 655 rectal cancer cases, as well as 2434 population controls. A self-administered questionnaire was used to collect information on cigarette smoking and other potential confounders including occupation, drinking water source, and dietary practices. Logistic regression models were used to calculate odds ratios (ORs) and 95% confidence intervals (CIs), after adjustment for age, total energy intake, and other site- and sex-specific confounders.

RESULTS

In both sexes, cigarette smoking (ever vs. never) was associated with risk of bladder cancer (OR = 2.5; 95% CI, 2.0-3.1 for males; OR = 2.7; 2.0-3.6 for females) and pancreatic cancer (OR = 1.8; 1.2-2.8 for males; OR = 2.1; 1.4-3.1 for females). Cigarette smoking also increased the risk of kidney cancer among males (OR = 1.8; 1.3-2.7), and to a lesser degree, among females (OR = 1.2; 0.8-1.8). No association was found for colon or rectal cancer in either sex.

CONCLUSIONS

Cigarette smoking increased the risk of bladder, kidney, and pancreatic cancer in men and women. The rankings of multivariate-adjusted ORs from highest to lowest were bladder, pancreas, kidney, and colorectum, with little difference between men and women.

Lifestyle and anthropometric risk factors for prostate cancer in a cohort of Iowa men

PURPOSE

Several lines of evidence suggest that prostate cancer has a hormonal etiology. We evaluated factors known to modulate the endocrine system, including alcohol and tobacco use, physical activity, and obesity as risk factors for prostate cancer.

METHODS

Cancer-free controls who participated in a population-based case-control study from 1986-1989 (81% response rate) were followed through 1995 for cancer incidence by linkage to the Iowa Cancer Registry; 101 incident prostate cancers were identified.

RESULTS

Compared with non-users of alcohol, men who consumed <22 grams alcohol per week (relative risk [RR] = 1.1; 95% Confidence Interval [CI] 0.6-2.1), 22-96 grams alcohol per week (RR = 2.6; 95% CI 1.4-4. 6) and >96 grams alcohol per week (RR = 3.1; 95% CI 1.5-6.3) were at increased risk of prostate cancer after adjustment for age, family history of prostate cancer, body mass index, total energy, and intake of carbohydrate, linoleic acid, lycopene, retinol, and red meat (p for trend < 0.0001). The respective RRs were similar when assessing type of alcohol consumed (beer, wine or liquor) or when well-differentiated, localized tumors were excluded. Body mass index was only weakly and positively associated with prostate cancer after adjustment for age, but this association strengthened after multivariate adjustment and exclusion of well-differentiated, localized tumors. For the latter tumors, men with a BMI of 24.1-26.6 kg/m(2) and >26.6 kg/m(2) were at elevated risk compared to men with a BMI <24.1 kg/m(2). Tobacco use (cigarettes, cigar/pipe, chewing tobacco and snuff use), height, weight, and both leisure and occupational physical activity were not associated with risk of prostate cancer in this cohort.

CONCLUSIONS

These data suggest that in white men obesity is a risk factor for more clinically significant prostate cancer and confirm limited previous reports showing that alcohol consumption is positively associated with prostate cancer and that this risk is not limited to any specific type of alcohol.

Design issues in case-control studies

The most difficult and most important considerations in planning the protocol of a case-control study are ascertainment of cases, selection of controls and the quality of the exposure measurement. Plans to ensure careful field work are equally important; without attention to data collection, the protocol will be meaningless. In most case-control studies, the measurement problem is magnified because one cannot implement the collection of exposure information at the beginning of follow-up, and instead must rely on interviews, existing records or extrapolation into the past. Consideration of a case-control study as an efficient way to study a cohort helps to resolve some design issues.

Response rates among control subjects in case-control studies

Response rates are an important component of epidemiologic research. The purposes of this study are (a) to evaluate how response rates are defined and calculated for control subjects in epidemiologic case-control studies, and (b) to explore factors that may impact response in epidemiologic studies. Our results show that the method of control subject selection has an impact on study response. Gender of respondent does not appear to impact response rates. However, response rates are generally worse for individuals less than 45 years old. Methods used to calculate response have a great impact on "response rate"; therefore, it is important for researchers to define exactly what the reported response rates represent and how they are derived so that data can be interpreted appropriately.

Incidence and risk factors for childhood brain tumors in the Ile de France

A case-control study investigating risk factors for childhood brain tumors was conducted in the Ile de France (Paris region). During a 2-year period (1985-1987) 109 newly diagnosed cases were identified and, of these, 75 could be interviewed. In the same region, 113 population controls, frequency-matched for year of birth, were interviewed. Odds ratios adjusted for child's age and sex and for maternal age were estimated for each risk factor present in utero or during childhood by conditional logistic regression. Statistically significant associations were found for the following risk factors: farm residence, cat scratches, home treated with pesticides, passive smoking, family history of cancer, antihistamine intake. Intake of vitamin supplements during childhood was associated with a decrease in risk. This study is part of a multicentric case-control study coordinated by the International Agency for Research on Cancer and its results will be compared for consistency, and pooled with those of other centers using the same protocol.

A comparison of three approaches to estimate exposure-specific incidence rates from population-based case-control data

In population-based case-control studies, an attempt is made to identify all incident cases diagnosed in a specified population during a fixed time interval. Assuming that this goal is met allows one to obtain measures of risk other than relative risks. In this paper, we describe three approaches to estimate exposure-specific incidence rates. Approach 1 relies on estimating crude incidence rates of the disease in strata defined, for instance, by age and geographic area, and combining them with relative risk estimates from the case-control data. In approaches 2 and 3, baseline incidence rates and relative risks are estimated jointly. Approach 2 is based on a pseudo-likelihood, while, in approach 3, the problem is regarded as a missing data problem and a full likelihood is maximized. We applied these three approaches to a study of bladder cancer. Our three sets of estimates of exposure-specific incidence rates were in close agreement, while there appeared to be greater precision with approaches 2 and 3.

The driver's license list as a population-based sampling frame in Iowa

Driver's license lists are infrequently used for population-based sampling, presumably because of suspicions of poor population coverage. The 1990 Iowa driver's license list was compared with the 1990 census to evaluate coverage by 5-year age group, sex, resident county, and urbanicity. Coverage exceeded 90% among 15- to 74-year-old men and 15- to 64-year-old women, with uniform coverage by county and county urbanicity group in these age ranges. In Iowa, these lists are convenient and cost-effective and appear to be representative for 25- to 64-year-olds. The representativeness of driver lists in regard to other factors and in other geographic regions deserves further evaluation.

Smoking and bladder cancer risk in blacks and whites in the United States

A population-based case-control study of bladder cancer (2,982 cases and 5,782 controls) conducted in 10 areas of the United States examined the effect of smoking as a risk factor among Blacks and Whites, after adjustment for occupation and other potential confounders. Although the overall risk for smoking was slightly higher in Blacks than Whites (relative risk = 2.7 and 2.2, respectively), this difference was not statistically significant. Estimation of risk by dose and currency of exposure revealed no consistent racial disparities in smoking-related risks. Race-specific, attributable risk estimates indicated that nearly half of bladder cancers among both Blacks and Whites could have been prevented by elimination of smoking.

Comparison of three inducement techniques to improve compliance in a health survey conducted by telephone

The use of telephone interviews for epidemiologic and public health studies has increased in recent years. Since telephone surveys are susceptible to lower response rates than personal interviews, several attempts have been reported to increase respondents' compliance using various precontact procedures. This investigation evaluates the comparative effectiveness of three techniques to enhance compliance with a relatively long telephone interview on epidemiologic topics. The theoretical and practical applications in the domain of telephone surveys of two techniques, the foot-in-the-door and the low ball, commonly considered nonpressure techniques, are discussed. A newly suggested, combined compliance procedure is also introduced and tested. Results show that compliance was greater for the new method when compared with each of the other two methods. Moreover, each of the three methods outperformed a control condition. The theoretical models developed to devise and explain the new techniques received empirical support in a public health survey employing 335 adult residents of Tel Aviv, Israel, in May 1988.

Smoking patterns by occupation and duration of employment

Lifetime patterns of smoking and occupation based on personal interviews were examined among 3,627 white men and 1,200 white women who were randomly selected from ten areas in the United States during the period 1977-1978. These individuals participated in the control series of the National Bladder Cancer Study. We estimated, based on Axelson's method, the extent to which smoking habits for given occupational groups would confound the estimated relative risk for lung cancer for 62 occupations among men and 18 occupations among women. Among men, confounding by smoking resulted in a 30% or greater increased risk of lung cancer in only three occupational groups--namely, stationary engineers and power station operators (relative risk (RR) = 1.6), printers (RR = 1.3), and fishermen and sailors (RR = 1.3). A decrease in lung cancer risk of 0.8 or less due to smoking habits was observed among the clergy (RR = 0.5) and chemical workers (RR = 0.7). Among women, a 30% increase or greater in the risk of lung cancer based on smoking habits alone was found for food service workers (RR = 1.5), building managers and administrators (RR = 1.3), telephone and telegraph operators (RR = 1.3), and operatives (RR = 1.3). A risk ratio of 0.8 or less was observed for those women employed as farmers (RR = 0.5) and teachers (RR = 0.8). Smoking habits by duration of employment were also examined for 38 occupations among men. The largest increase in the risk of lung cancer based on the smoking habits among long-term workers was only 1.3 and was observed for those men employed 20 or more years as painters and as electricians. These findings suggest that the smoking patterns, in only a few occupational groups that we evaluated, confound estimates of the relative risk by more than 30%, and for most occupational groups under investigation in this study, confounding by smoking alone did not produce trends in relative risks by duration of employment.

Chlorinated drinking water and bladder cancer: effect of misclassification on risk estimates

Data are presented from the Iowa portion of the National Bladder Cancer Case-Control Study demonstrating the effect of misclassification on depressing odds ratio estimates for years of exposure to chlorinated drinking water and bladder cancer. Four methods (METHODS 1 through 4) of quantifying chlorination exposure with sequentially decreasing degrees of misclassification are presented for the 268 bladder cancer cases and 658 population-based controls fulfilling criteria for inclusion in this study. Twenty-eight other risk factors for bladder cancer were considered along with chlorinated drinking water exposure estimated by METHOD 4. Stepwise regression models included as significant factors cigarette smoking (p less than .001), chlorination exposure (p = .038), and irradiation to the pelvic area (p = .040). Replacement of chlorinated drinking water exposure estimated by METHOD 4 with any of the remaining three methods resulted in models that included cigarette smoking and irradiation to the pelvic area, but not chlorination exposure. Thus, misclassification of chlorination exposure signified the difference between observing and not observing an association with bladder cancer.

Occupation and bladder cancer in Utah

The relationships between bladder cancer and occupation, industries, and occupational exposures in Utah were examined in a population-based, case-control study conducted between 1977 and 1983. Life-long occupational histories were obtained for 417 cases (332 men and 85 women) and 877 controls (685 men and 192 women). Although few positive findings emerged in this study, increased risks were detected among men for employment in the leather and textile industries which increased with duration of employment. The effects were most marked for employment beginning 45 or more years prior to interview (odds ratio [OR] for textiles = 1.92, confidence interval [CI] = 0.89-4.46; for leather OR = 2.95, CI = 0.63-13.76). Among men and women, increased risk was detected among clerical workers employed for less than 10 years (OR = 1.59, CI = 1.16-2.17) although the risk decreased with increased duration of employment (OR = 0.88, CI = 0.55-1.40 for greater than or equal to 10 years). A protective effect was seen among men and women for 10 or more years employment in professional, managerial, and technical occupations (OR = 0.68, CI = 0.50-0.92). Employment as a carpenter resulted in increased risk which increased with duration. Increased risk for bladder cancer was detected among carpenters who smoked but not among carpenters who never smoked. We used an occupation-exposure linkage system to identify workers exposed to aromatic amino compounds; such workers did not have increased risk of bladder cancer, although interaction between long-term exposure to aromatic amino compounds and smoking was detected. Interactions between smoking and other industrial or occupational exposures were not demonstrated, and for the most part, smoking did not confound the estimates of the bladder cancer-occupation relationships.

Fluid intake and bladder cancer in Utah

A population-based, incident case-control study was conducted in Utah to assess the relationship between fluid intake and bladder cancer. Cancer cases (n = 419) were identified through the Utah Cancer Registry, and controls (n = 889) were obtained through random digit dialing and the Health Care Financing Administration. After adjustment for cigarette smoking, age, sex, history of diabetes, and history of bladder infections using multiple logistic regression analysis, total fluid intake was not found to be related to bladder cancer development. Specific fluids related to bladder cancer risk were milk intake (OR = 0.64) and caffeinated coffee intake (OR = 1.60). A linear trend for a dose-response protective effect was observed for milk, while coffee increased risk only when 40 or more cups were consumed per week. Alcohol increased risk only when consumed at high levels (over 3.64 ounces or 103 g per week) by people who never smoked cigarettes (OR = 2.37). Likewise, tea consumption in non-cigarette smokers increased bladder cancer risk (OR = 2.25). Results from this study suggest that types of fluids consumed may play a role in the development of bladder cancer. Furthermore, it is hypothesized that the dietary components of these beverages may be related to the development of bladder cancer.

Direct methods of obtaining information on cigarette smoking in occupational studies

Many occupational epidemiology studies require complete and accurate information on tobacco use to control for confounding by smoking and to assess interactions of smoking with workplace exposures. This paper reviews and evaluates the availability, reliability, validity, and efficiency of the various data sources and techniques for obtaining individual smoking data, including existing records, biological markers, and surveys. Emphasis is placed on the highly problematic issue of obtaining retrospective smoking histories. In general, the survey technique is currently deemed the most feasible approach for obtaining lifetime smoking histories. Both theoretical and practical aspects of smoking surveys are discussed in detail and are illustrated with a review of the recent literature and with data from two recent retrospective cohort studies conducted at the University of Pittsburgh. Several recommendations involving both the use of smoking data and areas for future methodologic research are presented. These include (1) justification for collecting smoking data in occupational studies based primarily on the potential for smoking to act as an effect modifier rather than solely as a confounder, (2) checks for reliability and validity in all studies which involve the collection of smoking data, (3) more methodologic research to better understand the impact that missing, unreliable, and invalid smoking data may have on the ability to detect and quantify important smoking-exposure interactions, and (4) an assessment of the correlation between biological markers and cigarette carcinogen exposure.

Tobacco and occupation as risk factors in bladder cancer: a case-control study in southern Belgium

A pilot case-control study on bladder cancer, with population-based controls matched for each of 74 cases, has been conducted in two industrial areas of Southern Belgium in order to analyze the influence of tobacco use and occupation. Observed bladder cancer risk for smokers is more than 5 times higher than that for non-smokers, and the risk for people having an a priori hazardous occupation is about 3.5 times higher than that of other subjects. A dose-response relationship was found for tobacco exposure and duration of employment. It seems that the risk is increased in a log-linear way by these variables. Population attributable risks show that in 20 cases of bladder cancer, 17 could be explained by the two factors combined. This study also reveals an increased risk for metal workers, truck and engine drivers, coal-miners, and rubber and coal-tar workers. The risk for metal workers is specially high in the case of turners, metal fitters, blacksmiths, stokers and workers exposed to hot metal.

Bladder cancer and occupational exposure to leather

A large case-control study of bladder cancer (2982 cases; 5782 controls) included information about occupational exposure to leather. Occupational histories of exposed white study subjects were reviewed and 150 were determined to have had "true" on the job exposure to leather. The odds ratio estimate (OR) of bladder cancer associated with such exposure in white subjects (n = 8063) was 1.4 (95% confidence limits = 1.0, 1.9) after adjustment for sex, age, and cigarette smoking. The risk was highest in those first employed in a leather job before 1945, although no dose-response relation with duration of leather employment was found. Subjects employed in "dusty" leather jobs had a slightly higher risk than those with other types of leather jobs. Our results are consistent with reports of an increased risk of bladder cancer associated with exposure to leather. Although the agents responsible have not been identified, our findings of an increased risk associated with exposure in the earlier years of this century and in dusty jobs suggest that leather dusts may be important.

The use of interviews with surrogate respondents in a case-control study of oral cancer

The purpose of this study was to assess the possible bias that may occur in case-control studies when exposure information is not collected from all potentially eligible cases. The data used in this study were collected in the metropolitan area of Atlanta as part of a multicenter, population-based, case-control study of oropharyngeal cancer. In-person interviews were conducted with 112 cases (67.9 percent) and information on an additional 23 ill or deceased cases (13.9 percent) was collected through surrogate respondents. The cases about whom information was collected from surrogate respondents had more advanced disease at the time of diagnosis and were more likely to be black and less educated than cases who were interviewed in person. Cigarette smoking and consumption of hard liquor were more common among the cases about whom information was collected through surrogates. Therefore, failure to include such information would have resulted in underestimates of the strength of association between these exposures and the risk of oropharyngeal cancer.

Cancer risk among artistic painters

An association between bladder cancer and employment as an artistic painter was found in two study populations. A proportionate mortality analysis of death certificates of professional artists indicated a greater than twofold excess for bladder cancer mortality among painters. This association was further investigated in a large case-control interview study of bladder cancer patients where an overall relative risk estimate of 2.5 was found among artistic painters. Excess deaths from leukemia and arteriosclerotic heart disease also occurred in the proportionate mortality study. Information was not available to determine exposure to specific substances that may have been responsible for these observations; however, efforts should be made to limit exposure to art materials known to be hazardous.

Determinants of medical record access for an epidemiologic study

Hospital record access is often needed for epidemiologic research. In a recent study, 26 Denver-area hospitals were contacted for permission to review records. The review process, the decision regarding access, and hospital characteristics were analyzed. Hospitals with a review committee, larger hospitals, and those which reached a decision rapidly were most likely to grant access. These results may provide guidance to investigators planning studies requiring review of hospital records.

Bladder cancer risk among auto and truck mechanics and chemically related occupations

Male participants in the NCI National Bladder Cancer study were evaluated for risks in auto and truck mechanic and in chemically related exposure occupations compared with those who had never worked in these areas. In nonsmokers the adjusted rate ratios (RR) for mechanics equalled 1.33, (95% CI 0.77, 2.31). The adjusted RR in the nonsmoker chemically related group was 1.53 (95% CI 1.13, 2.07). Duration of exposure as a mechanic was not clearly associated with bladder cancer risk regardless of smoking status.

Causes of death among female chemists

A mortality odds ratio analysis of cause of death among 347 White female members of the American Chemical Society (ACS) revealed a five-fold excess of suicide, notably by cyanide poisoning. Risk was also elevated for all cancers combined and for cancers of the breast, ovary, stomach, pancreas, and lymphatic and hematopoietic system. The excess breast and ovary cancer deaths were limited to unmarried women.

Familial and environmental interactions in bladder cancer risk

In a population-based study of 2,982 bladder cancer patients and 5,782 controls in 10 geographic areas of the United States which was designed to assess the role of environmental risk factors, information was also obtained on the history of urinary tract cancer in first-degree relatives. A family history of urinary tract cancer significantly elevated the risk of bladder cancer [relative risk (RR) = 1.45], with higher risks observed among patients under age 45. The risks of bladder cancer associated with positive family history were generally higher among persons with suspected environmental exposures, particularly heavy cigarette smoking (RR = 10.7 among those who smoked 3 or more packs per day). Further studies of bladder cancer should incorporate biochemical and genetic probes to assess mechanisms of familial susceptibility and interactions with environmental factors.

Bladder cancer risk and pipes, cigars, and smokeless tobacco

Interview data from 2982 patients with bladder cancer and 5782 controls selected from the general population were used to assess the effects of non-cigarette tobacco use on bladder cancer risk. Compared to men who had never smoked, those who had smoked pipes but not cigars or cigarettes had a relative risk estimated at 1.23 (95% confidence interval [CI] = 0.75-2.00). Those who smoked cigars but not pipes or cigarettes were estimated to have a relative risk of 1.33 (95% CI = 0.92-1.94). Little evidence of dose response was observed. The excess relative risk to pipe smokers was limited to those who inhaled deeply.